Structure-guided discovery of protein and glycan components in native mastigonemes.

Mastigonemes, the hair-like lateral appendages lining cilia or flagella, participate in mechanosensation and cellular motion, but their constituents and structure have remained unclear. Here, we report the cryo-EM structure of native mastigonemes isolated from Chlamydomonas at 3.0 Å resolution. The long stem assembles as a super spiral, with each helical turn comprising four pairs of anti-parallel mastigoneme-like protein 1 (Mst1). A large array of arabinoglycans, which represents a common class of glycosylation in plants and algae, is resolved surrounding the type II poly-hydroxyproline (Hyp) helix in Mst1. The EM map unveils a mastigoneme axial protein (Mstax) that is rich in heavily glycosylated Hyp and contains a PKD2-like transmembrane domain (TMD). Mstax, with nearly 8,000 residues spanning from the intracellular region to the distal end of the mastigoneme, provides the framework for Mst1 assembly. Our study provides insights into the complexity of protein and glycan interactions in native bio-architectures.

Osmosensor-mediated control of Ca2+ spiking in pollen germination.

Higher plants survive terrestrial water deficiency and fluctuation by arresting cellular activities (dehydration) and resuscitating processes (rehydration). However, how plants monitor water availability during rehydration is unknown. Although increases in hypo-osmolarity-induced cytosolic Ca2+ concentration (HOSCA) have long been postulated to be the mechanism for sensing hypo-osmolarity in rehydration1,2, the molecular basis remains unknown. Because osmolarity triggers membrane tension and the osmosensing specificity of osmosensing channels can only be determined in vivo3-5, these channels have been classified as a subtype of mechanosensors. Here we identify bona fide cell surface hypo-osmosensors in Arabidopsis and find that pollen Ca2+ spiking is controlled directly by water through these hypo-osmosensors-that is, Ca2+ spiking is the second messenger for water status. We developed a functional expression screen in Escherichia coli for hypo-osmosensitive channels and identified OSCA2.1, a member of the hyperosmolarity-gated calcium-permeable channel (OSCA) family of proteins6. We screened single and high-order OSCA mutants, and observed that the osca2.1/osca2.2 double-knockout mutant was impaired in pollen germination and HOSCA. OSCA2.1 and OSCA2.2 function as hypo-osmosensitive Ca2+-permeable channels in planta and in HEK293 cells. Decreasing osmolarity of the medium enhanced pollen Ca2+ oscillations, which were mediated by OSCA2.1 and OSCA2.2 and required for germination. OSCA2.1 and OSCA2.2 convert extracellular water status into Ca2+ spiking in pollen and may serve as essential hypo-osmosensors for tracking rehydration in plants.

Genome-wide RNA structure changes during human neurogenesis modulate gene regulatory networks.

The distribution, dynamics, and function of RNA structures in human development are under-explored. Here, we systematically assayed RNA structural dynamics and their relationship with gene expression, translation, and decay during human neurogenesis. We observed that the human ESC transcriptome is globally more structurally accessible than differentiated cells and undergoes extensive RNA structure changes, particularly in the 3' UTR. Additionally, RNA structure changes during differentiation are associated with translation and decay. We observed that RBP and miRNA binding is associated with RNA structural changes during early neuronal differentiation, and splicing is associated during later neuronal differentiation. Furthermore, our analysis suggests that RBPs are major factors in structure remodeling and co-regulate additional RBPs and miRNAs through structure. We demonstrated an example of this by showing that PUM2-induced structure changes on LIN28A enable miR-30 binding. This study deepens our understanding of the widespread and complex role of RNA-based gene regulation during human development.

Vertical MoS2 transistors with sub-1-nm gate lengths.

Ultra-scaled transistors are of interest in the development of next-generation electronic devices1-3. Although atomically thin molybdenum disulfide (MoS2) transistors have been reported4, the fabrication of devices with gate lengths below 1 nm has been challenging5. Here we demonstrate side-wall MoS2 transistors with an atomically thin channel and a physical gate length of sub-1 nm using the edge of a graphene layer as the gate electrode. The approach uses large-area graphene and MoS2 films grown by chemical vapour deposition for the fabrication of side-wall transistors on a 2-inch wafer. These devices have On/Off ratios up to 1.02 × 105 and subthreshold swing values down to 117 mV dec-1. Simulation results indicate that the MoS2 side-wall effective channel length approaches 0.34 nm in the On state and 4.54 nm in the Off state. This work can promote Moore's law of the scaling down of transistors for next-generation electronics.

CCR5 closes the temporal window for memory linking.

Real-world memories are formed in a particular context and are often not acquired or recalled in isolation1-5. Time is a key variable in the organization of memories, as events that are experienced close in time are more likely to be meaningfully associated, whereas those that are experienced with a longer interval are not1-4. How the brain segregates events that are temporally distinct is unclear. Here we show that a delayed (12-24 h) increase in the expression of C-C chemokine receptor type 5 (CCR5)-an immune receptor that is well known as a co-receptor for HIV infection6,7-after the formation of a contextual memory determines the duration of the temporal window for associating or linking that memory with subsequent memories. This delayed expression of CCR5 in mouse dorsal CA1 neurons results in a decrease in neuronal excitability, which in turn negatively regulates neuronal memory allocation, thus reducing the overlap between dorsal CA1 memory ensembles. Lowering this overlap affects the ability of one memory to trigger the recall of the other, and therefore closes the temporal window for memory linking. Our findings also show that an age-related increase in the neuronal expression of CCR5 and its ligand CCL5 leads to impairments in memory linking in aged mice, which could be reversed with a Ccr5 knockout and a drug approved by the US Food and Drug Administration (FDA) that inhibits this receptor, a result with clinical implications. Altogether, the findings reported here provide insights into the molecular and cellular mechanisms that shape the temporal window for memory linking.

Aging and comprehensive molecular profiling in acute myeloid leukemia.

Acute myeloid leukemia (AML) is an aging-related and heterogeneous hematopoietic malignancy. In this study, a total of 1,474 newly diagnosed AML patients with RNA sequencing data were enrolled, and targeted or whole exome sequencing data were obtained in 94% cases. The correlation of aging-related factors including age and clonal hematopoiesis (CH), gender, and genomic/transcriptomic profiles (gene fusions, genetic mutations, and gene expression networks or pathways) was systematically analyzed. Overall, AML patients aged 60 y and older showed an apparently dismal prognosis. Alongside age, the frequency of gene fusions defined in the World Health Organization classification decreased, while the positive rate of gene mutations, especially CH-related ones, increased. Additionally, the number of genetic mutations was higher in gene fusion-negative (GF-) patients than those with GF. Based on the status of CH- and myelodysplastic syndromes (MDS)-related mutations, three mutant subgroups were identified among the GF- AML cohort, namely, CH-AML, CH-MDS-AML, and other GF- AML. Notably, CH-MDS-AML demonstrated a predominance of elderly and male cases, cytopenia, and significantly adverse clinical outcomes. Besides, gene expression networks including HOXA/B, platelet factors, and inflammatory responses were most striking features associated with aging and poor prognosis in AML. Our work has thus unraveled the intricate regulatory circuitry of interactions among different age, gender, and molecular groups of AML.

STMHCpan, an accurate Star-Transformer-based extensible framework for predicting MHC I allele binding peptides.

Peptide-major histocompatibility complex I (MHC I) binding affinity prediction is crucial for vaccine development, but existing methods face limitations such as small datasets, model overfitting due to excessive parameters and suboptimal performance. Here, we present STMHCPan (STAR-MHCPan), an open-source package based on the Star-Transformer model, for MHC I binding peptide prediction. Our approach introduces an attention mechanism to improve the deep learning network architecture and performance in antigen prediction. Compared with classical deep learning algorithms, STMHCPan exhibits improved performance with fewer parameters in receptor affinity training. Furthermore, STMHCPan outperforms existing ligand benchmark datasets identified by mass spectrometry. It can also handle peptides of arbitrary length and is highly scalable for predicting T-cell responses. Our software is freely available for use, training and extension through Github (https://github.com/Luckysoutheast/STMHCPan.git).

Auditory training remodels hippocampus-related memory in adult rats.

Consequences of perceptual training, such as improvements in discriminative ability, are highly stimulus and task specific. Therefore, most studies on auditory training-induced plasticity in adult brain have focused on the sensory aspects, particularly on functional and structural effects in the auditory cortex. Auditory training often involves, other than auditory demands, significant cognitive components. Yet, how auditory training affects cognition-related brain regions, such as the hippocampus, remains unclear. Here, we found in female rats that auditory cue-based go/no-go training significantly improved the memory-guided behaviors associated with hippocampus. The long-term potentiations of the trained rats recorded in vivo in the hippocampus were also enhanced compared with the naïve rats. In parallel, the phosphorylation level of calcium/calmodulin-dependent protein kinase II and the expression of parvalbumin-positive interneurons in the hippocampus were both upregulated. These findings demonstrate that auditory training substantially remodels the processing and function of brain regions beyond the auditory system, which are associated with task demands.

Hybrid measurement of respiratory aerosol reveals a dominant coarse fraction resulting from speech that remains airborne for minutes.

Accurate measurements of the size and quantity of aerosols generated by various human activities in different environments are required for efficacious mitigation strategies and accurate modeling of respiratory disease transmission. Previous studies of speech droplets, using standard aerosol instrumentation, reported very few particles larger than 5 µm. This starkly contrasts with the abundance of such particles seen in both historical slide deposition measurements and more recent light scattering observations. We have reconciled this discrepancy by developing an alternative experimental approach that addresses complications arising from nucleated condensation. Measurements reveal that a large volume fraction of speech-generated aerosol has diameters in the 5- to 20-µm range, making them sufficiently small to remain airborne for minutes, not hours. This coarse aerosol is too large to penetrate the lower respiratory tract directly, and its relevance to disease transmission is consistent with the vast majority of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infections initiating in the upper respiratory tract. Our measurements suggest that in the absence of symptoms such as coughing or sneezing, the importance of speech-generated aerosol in the transmission of respiratory diseases is far greater than generally recognized.

Transcriptome-based molecular subtypes and differentiation hierarchies improve the classification framework of acute myeloid leukemia.

The current classification of acute myeloid leukemia (AML) relies largely on genomic alterations. Robust identification of clinically and biologically relevant molecular subtypes from nongenomic high-throughput sequencing data remains challenging. We established the largest multicenter AML cohort (n = 655) in China, with all patients subjected to RNA sequencing (RNA-Seq) and 619 (94.5%) to targeted or whole-exome sequencing (TES/WES). Based on an enhanced consensus clustering, eight stable gene expression subgroups (G1-G8) with unique clinical and biological significance were identified, including two unreported (G5 and G8) and three redefined ones (G4, G6, and G7). Apart from four well-known low-risk subgroups including PML::RARA (G1), CBFB::MYH11 (G2), RUNX1::RUNX1T1 (G3), biallelic CEBPA mutations or -like (G4), four meta-subgroups with poor outcomes were recognized. The G5 (myelodysplasia-related/-like) subgroup enriched clinical, cytogenetic and genetic features mimicking secondary AML, and hotspot mutations of IKZF1 (p.N159S) (n = 7). In contrast, most NPM1 mutations and KMT2A and NUP98 fusions clustered into G6-G8, showing high expression of HOXA/B genes and diverse differentiation stages, from hematopoietic stem/progenitor cell down to monocyte, namely HOX-primitive (G7), HOX-mixed (G8), and HOX-committed (G6). Through constructing prediction models, the eight gene expression subgroups could be reproduced in the Cancer Genome Atlas (TCGA) and Beat AML cohorts. Each subgroup was associated with distinct prognosis and drug sensitivities, supporting the clinical applicability of this transcriptome-based classification of AML. These molecular subgroups illuminate the complex molecular network of AML, which may promote systematic studies of disease pathogenesis and foster the screening of targeted agents based on omics.

Genomic and transcriptomic characteristics of 12 novel primary cell lines derived from three patients with cholangiocarcinoma.

Cholangiocarcinoma (CCA) is an aggressive bile duct malignancy with poor prognosis. To improve our understanding of the biological characteristics of CCA and develop effective therapies, appropriate preclinical models are required. Here, we established and characterized 12 novel patient-derived primary cancer cell (PDPC) models using multi-region sampling. At the genomic level of PDPCs, we observed not only commonly mutated genes, such as TP53, JAK3, and KMT2C, consistent with the reports in CCA, but also specific mutation patterns in each cell line. In addition, specific expression patterns with distinct biological functions and pathways involved were also observed in the PDPCs at the transcriptomic level. Furthermore, the drug-sensitivity results revealed that the PDPCs exhibited different responses to the six commonly used compounds. Our findings indicate that the established PDPCs can serve as novel in vitro reliable models to provide a crucial molecular basis for improving the understanding of tumorigenesis and its treatment.

MST1 knockdown inhibits osteoarthritis progression through Parkin-mediated mitophagy and Nrf2/NF-κB signalling pathway.

Osteoarthritis (OA) is a complicated disease that involves apoptosis and mitophagy. MST1 is a pro-apoptotic factor. Hence, decreasing its expression plays an anti-apoptotic effect. This study aims to investigate the protective effect of MST1 inhibition on OA and the underlying processes. Immunofluorescence (IF) was used to detect MST1 expression in cartilage tissue. Western Blot, ELISA and IF were used to analyse the expression of inflammation, extracellular matrix (ECM) degradation, apoptosis and mitophagy-associated proteins. MST1 expression in chondrocytes was inhibited using siRNA and shRNA in vitro and in vivo. Haematoxylin-Eosin, Safranin O-Fast Green and alcian blue staining were used to evaluate the therapeutic effect of inhibiting MST1. This study discovered that the expression of MST1 was higher in OA patients. Inhibition of MST1 reduced inflammation, ECM degradation and apoptosis and enhanced mitophagy in vitro. MST1 inhibition slows OA progression in vivo. Inhibiting MST1 suppressed apoptosis, inflammation and ECM degradation via promoting Parkin-mediated mitophagy and the Nrf2-NF-κB axis. The results suggest that MST1 is a possible therapeutic target for the treatment of osteoarthritis as its inhibition delays the progression of OA through the Nrf2-NF-κB axis and mitophagy.

Hemodynamic force dictates endothelial angiogenesis through MIEN1-ERK/MAPK-signaling axis.

It is well-recognized that blood flow at branches and bends of arteries generates disturbed shear stress, which plays a crucial in driving atherosclerosis. Flow-generated fluid shear stress (FSS), as one of the key hemodynamic factors, is appreciated for its critical involvement in regulating angiogenesis to facilitate wound healing and tissue repair. Endothelial cells can directly sense FSS but the mechanobiological mechanism by which they decode different patterns of FSS to trigger angiogenesis remains unclear. In the current study, laminar shear stress (LSS, 15 dyn/cm2) was employed to mimic physiological blood flow, while disturbed shear stress (DSS, ranging from 0.5 ± 4 dyn/cm2) was applied to simulate pathological conditions. The aim was to investigate how these distinct types of blood flow regulated endothelial angiogenesis. Initially, we observed that DSS impaired angiogenesis and downregulated endogenous vascular endothelial growth factor B (VEGFB) expression compared to LSS. We further found that the changes in membrane protein, migration and invasion enhancer 1 (MIEN1) play a role in regulating ERK/MAPK signaling, thereby contributing to endothelial angiogenesis in response to FSS. We also showed the involvement of MIEN1-directed cytoskeleton organization. These findings suggest the significance of shear stress in endothelial angiogenesis, thereby enhancing our understanding of the alterations in angiogenesis that occur during the transition from physiological to pathological blood flow.

Quantitative Characterization of Chain-Flipping of Acyl Carrier Protein of Escherichia coli Using Chemical Exchange NMR.

The acyl carrier protein of Escherichia coli, termed AcpP, is a prototypical example of type II fatty acid synthase systems found in many bacteria. It serves as a central hub by accepting diverse acyl moieties (4-18 carbons) and shuttling them between its multiple enzymatic partners to generate fatty acids. Prior structures of acyl-AcpPs established that thioester-linked acyl cargos are sequestered within AcpP's hydrophobic lumen. In contrast, structures of enzyme-bound acyl-AcpPs showed translocation of AcpP-tethered acyl chains into the active sites of enzymes. The mechanistic underpinnings of this conformational interplay, termed chain-flipping, are unclear. Here, using heteronuclear NMR spectroscopy, we reveal that AcpP-tethered acyl chains (6-10 carbons) spontaneously adopt lowly populated solvent-exposed conformations. To this end, we devised a new strategy to replace AcpP's thioester linkages with 15N-labeled amide bonds, which facilitated direct "visualization" of these excited states using NMR chemical exchange saturation transfer and relaxation dispersion measurements. Global fitting of the corresponding data yielded kinetic rate constants of the underlying equilibrium and populations and lifetimes of solvent-exposed states. The latter were influenced by acyl chain composition and ranged from milliseconds to submilliseconds for chains containing six, eight, and ten carbons, owing to their variable interactions with AcpP's hydrophobic core. Although transient, the exposure of AcpP-tethered acyl chains to the solvent may allow relevant enzymes to gain access to its active thioester, and the enzyme-induced selection of this conformation will culminate in the production of fatty acids.

Structure-informed protein language models are robust predictors for variant effects.

Emerging variant effect predictors, protein language models (pLMs) learn evolutionary distribution of functional sequences to capture fitness landscape. Considering that variant effects are manifested through biological contexts beyond sequence (such as structure), we first assess how much structure context is learned in sequence-only pLMs and affecting variant effect prediction. And we establish a need to inject into pLMs protein structural context purposely and controllably. We thus introduce a framework of structure-informed pLMs (SI-pLMs), by extending masked sequence denoising to cross-modality denoising for both sequence and structure. Numerical results over deep mutagenesis scanning benchmarks show that our SI-pLMs, even when using smaller models and less data, are robustly top performers against competing methods including other pLMs, which shows that introducing biological context can be more effective at capturing fitness landscape than simply using larger models or bigger data. Case studies reveal that, compared to sequence-only pLMs, SI-pLMs can be better at capturing fitness landscape because (a) learned embeddings of low/high-fitness sequences can be more separable and (b) learned amino-acid distributions of functionally and evolutionarily conserved residues can be of much lower entropy, thus much more conserved, than other residues. Our SI-pLMs are applicable to revising any sequence-only pLMs through model architecture and training objectives. They do not require structure data as model inputs for variant effect prediction and only use structures as context provider and model regularizer during training.

Assessing predictions on fitness effects of missense variants in HMBS in CAGI6.

This paper presents an evaluation of predictions submitted for the "HMBS" challenge, a component of the sixth round of the Critical Assessment of Genome Interpretation held in 2021. The challenge required participants to predict the effects of missense variants of the human HMBS gene on yeast growth. The HMBS enzyme, critical for the biosynthesis of heme in eukaryotic cells, is highly conserved among eukaryotes. Despite the application of a variety of algorithms and methods, the performance of predictors was relatively similar, with Kendall's tau correlation coefficients between predictions and experimental scores around 0.3 for a majority of submissions. Notably, the median correlation (≥ 0.34) observed among these predictors, especially the top predictions from different groups, was greater than the correlation observed between their predictions and the actual experimental results. Most predictors were moderately successful in distinguishing between deleterious and benign variants, as evidenced by an area under the receiver operating characteristic (ROC) curve (AUC) of approximately 0.7 respectively. Compared with the recent two rounds of CAGI competitions, we noticed more predictors outperformed the baseline predictor, which is solely based on the amino acid frequencies. Nevertheless, the overall accuracy of predictions is still far short of positive control, which is derived from experimental scores, indicating the necessity for considerable improvements in the field. The most inaccurately predicted variants in this round were associated with the insertion loop, which is absent in many orthologs, suggesting the predictors still heavily rely on the information from multiple sequence alignment.

In Situ Complexation of sgRNA and Cas12a Improves the Performance of a One-Pot RPA-CRISPR-Cas12 Assay.

Due to their ability to selectively target pathogen-specific nucleic acids, CRISPR-Cas systems are increasingly being employed as diagnostic tools. "One-pot" assays that combine nucleic acid amplification and CRISPR-Cas systems (NAAT-CRISPR-Cas) in a single step have emerged as one of the most popular CRISPR-Cas biosensing formats. However, operational simplicity comes at a cost, with one-pot assays typically being less sensitive than corresponding two-step NAAT-CRISPR-Cas assays and often failing to detect targets at low concentrations. It is thought that these performance reductions result from the competition between the two enzymatic processes driving the assay, namely, Cas-mediated cis-cleavage and polymerase-mediated amplification of the target DNA. Herein, we describe a novel one-pot RPA-Cas12a assay that circumvents this issue by leveraging in situ complexation of the target-specific sgRNA and Cas12a to purposefully limit the concentration of active Cas12a during the early stages of the assay. Using a clinically relevant assay against a DNA target for HPV-16, we show how this in situ format reduces competition between target cleavage and amplification and engenders significant improvements in detection limit when compared to the traditional one-pot assay format, even in patient-derived samples. Finally, to gain further insight into the assay, we use experimental data to formulate a mechanistic model describing the competition between the Cas enzyme and nucleic acid amplification. These findings suggest that purposefully limiting cis-cleavage rates of Cas proteins is a viable strategy for improving the performance of one-pot NAAT-CRISPR-Cas assays.

Comprehensive Crystal Regulation Reduces Interfacial Energy Loss for Efficient Blue Perovskite Light-Emitting Diodes.

As an essential component of future full-color displays, blue perovskite light-emitting diodes (PeLEDs) still lag far behind the red and green counterparts in the device performances. In the mainstream quasi-2D blue perovskite system, trap-mediated nonradiative loss, low energy transfer efficiency, and interface fluorescence quenching remain significant challenges. Herein, guanidinium thiocyanate (GASCN) and potassium cinnamate (PCA) are respectively introduced into the hole transport layer (HTL) and the perovskite precursor to achieve a dense and uniform perovskite thin film with greatly improved optoelectronic properties. Therefore, adequate GA+ acts as pre-nucleation sites on the HTL surface, regulating crystallization through strong hydrogen bonding with perovskite intermediates. The realized polydisperse domain distribution is conducive to cascade energy transfer, and the improved hole transport ability alleviates interface fluorescence quenching. In addition, the SCN- and CA- groups can form coordination bonds with the defects at the buried perovskite interface and grain boundaries, respectively, which effectively suppresses the detrimental nonradiative recombination. Benefitting from the comprehensive crystal regulation, blue PeLEDs featuring stable emission at 484 and 468 nm exhibit improved external quantum efficiencies of 11.5% and 4.3%, respectively.

Hiplot: a comprehensive and easy-to-use web service for boosting publication-ready biomedical data visualization.

Complex biomedical data generated during clinical, omics and mechanism-based experiments have increasingly been exploited through cloud- and visualization-based data mining techniques. However, the scientific community still lacks an easy-to-use web service for the comprehensive visualization of biomedical data, particularly high-quality and publication-ready graphics that allow easy scaling and updatability according to user demands. Therefore, we propose a community-driven modern web service, Hiplot (https://hiplot.org), with concise and top-quality data visualization applications for the life sciences and biomedical fields. This web service permits users to conveniently and interactively complete a few specialized visualization tasks that previously could only be conducted by senior bioinformatics or biostatistics researchers. It covers most of the daily demands of biomedical researchers with its equipped 240+ biomedical data visualization functions, involving basic statistics, multi-omics, regression, clustering, dimensional reduction, meta-analysis, survival analysis, risk modelling, etc. Moreover, to improve the efficiency in use and development of plugins, we introduced some core advantages on the client-/server-side of the website, such as spreadsheet-based data importing, cross-platform command-line controller (Hctl), multi-user plumber workers, JavaScript Object Notation-based plugin system, easy data/parameters, results and errors reproduction and real-time updates mode. Meanwhile, using demo/real data sets and benchmark tests, we explored statistical parameters, cancer genomic landscapes, disease risk factors and the performance of website based on selected native plugins. The statistics of visits and user numbers could further reflect the potential impact of this web service on relevant fields. Thus, researchers devoted to life and data sciences would benefit from this emerging and free web service.

Multimodal learning of noncoding variant effects using genome sequence and chromatin structure.

MOTIVATION: A growing amount of noncoding genetic variants, including single-nucleotide polymorphisms, are found to be associated with complex human traits and diseases. Their mechanistic interpretation is relatively limited and can use the help from computational prediction of their effects on epigenetic profiles. However, current models often focus on local, 1D genome sequence determinants and disregard global, 3D chromatin structure that critically affects epigenetic events. RESULTS: We find that noncoding variants of unexpected high similarity in epigenetic profiles, with regards to their relatively low similarity in local sequences, can be largely attributed to their proximity in chromatin structure. Accordingly, we have developed a multimodal deep learning scheme that incorporates both data of 1D genome sequence and 3D chromatin structure for predicting noncoding variant effects. Specifically, we have integrated convolutional and recurrent neural networks for sequence embedding and graph neural networks for structure embedding despite the resolution gap between the two types of data, while utilizing recent DNA language models. Numerical results show that our models outperform competing sequence-only models in predicting epigenetic profiles and their use of long-range interactions complement sequence-only models in extracting regulatory motifs. They prove to be excellent predictors for noncoding variant effects in gene expression and pathogenicity, whether in unsupervised "zero-shot" learning or supervised "few-shot" learning. AVAILABILITY AND IMPLEMENTATION: Codes and data can be accessed at https://github.com/Shen-Lab/ncVarPred-1D3D and https://zenodo.org/record/7975777.